摘要
贝叶斯网络被认为是人工智能研究中不确定性知识表示和推理的重要工具。当前在系统安全领域中已开始运用贝叶斯网络技术进行故障诊断分析,然而故障只是诱发事故的因素之一,无法系统的评价事故背后的隐患,对事故后果的预测也甚少涉及。笔者将贝叶斯网络作为一种事故分析手段,在事故致因理论的基础上提出了一种基于危险因素—事故—事故危害的三层贝叶斯网络拓扑模型;阐述了网络模型层次间的因果关联关系、各层次的构成、节点的描述方法以及网络模型的构建方法;最后通过一个天然气球罐的分析案例验证了该模型分析方法的可行性和有效性。
Bayesian network is considered to be an important method for the expression and reasoning of uncertainty in artificial intelligence research field. At present Bayesian network is applied to diagnose faults in system safety domain, but as one of various types of contributing causes, fault diagnosis is unable to evaluate hidden weaknesses systemically behind accidents of which the harmful effect is seldom considered as well. Therefore with Bayesian network being an approach of accident analysis, a three- layered Bayesian network topological model is proposed on the basis of risk factor- accident - accident hazards according to accident - causing theory. Moreover the linked relationships between the layers, the constituent elements and their descriptions at each layer and the model construction are expounded. Finally the feasibility and effectiveness of this accident analysis method is illuminated through a practical analysis example of a natural tins spherical tank.
出处
《中国安全生产科学技术》
CAS
2006年第4期45-50,共6页
Journal of Safety Science and Technology
关键词
系统安全
事故分析
贝叶斯网络
事故致因理论
system safety
accident analysis
bayesian network
accident - causing theory
